Is there any additional guide for custom Ray Container and minimal library requirements? This is for a kubernetes environment.
For a little bit of context, in a corporate environment I have to build a custom Ray container, I can’t use rayproject/ray
images. On top of that I have to build using our internal rhel image as base and its RPMs (or make the adequate software request to embed into the container).
So, my question is, is there a list for the absolute minimal system packages required to a ray[all]
installation in container to work? (Or ray[subsets])
Fortunately the python packages are easier to work, so not worried about them, only had to get rid of Anaconda. (but knowing which ones I don’t need is also handy)
So far I have based myself on this ray’s github dockerfile.
In summary, of these libraries, which are absolute necessary or which functionality I’d be giving up not installing any of them?
tzdata \
git \
libjemalloc-dev \
wget \
cmake \
g++ \
zlib1g-dev \
$(if [ "$AUTOSCALER" = "autoscaler" ]; then echo \
tmux \
screen \
rsync \
netbase \
openssh-client \
gnupg
I had some testing which I could remove some of them, but not really confident. Might just be a functionality that I am yet to use.